My help
The guide is good for basics but doesn't cover any of the NF7's issues.
1. The missing 10/10.5x multipliers over 225 MHz FSB with CPU-Interface enabled.
2. The 220 FSB wall with a tbred
3. CPU Interface issues
4. Differenent "better" bios's
5. Vcore / vdimm and +3.3v rail requirements
6. How long to run prime and what to look for
7. How to recover from potential non POST'ing
8. Memory/timing and restrictions.
Seen as im overly familair with the NF7 i'll try to clarify some of these things rather than just critising.
Again please note:
I AM NOT RESPONSIBLE FOR ANY PROBLEMS YOU ENCOUNTER FROM USING MY HELP.
All works fine for me so you shouldn't have any major problems.
---------------------------------------------------------------------------------
POINT 1 - The missing 10/10.5x multipliers over 225 MHz FSB with CPU-Interface enabled.
I cover CPU interface in more detail in Point 3.
On "most" NF7/S boards when you go over 225 MHz FSB with a 10 or 10.5x multiplier the system will refuse to post. This is only fixable if you use bios d_10.bin, this is the only bios with this problem does not occur. To use d_10 you must use the windows based flash utility. The bios is perfectly stable on most NF7's but unfortunately the SATA bios in it has corruption problems. There are variants of D_10.bin with the latest sata bios so you can search round for these.
Ok firstly using the basic guide find your CPU limits and use the vcore voltage it requires to do 2260 MHz, if you cant do 2260 Mhz you dont need to worry about this issue so move on.
Then on the bios you are using set as follows, in this example im using an week 28, 2800+ AQUCA which does 2260 MHz prime stable with 1.65v:
Softmenu III Setup:
CPU operating speed > user define
- External Clock > 226 MHz
- Multiplier Factor > x10
AGP Frequency > 66 MHz
CPU FSB/DRAM ratio ratio - 6/6
CPU Interface - enabled
Power Supply Controller - user define
- CPU Core Voltage >
1.675v
- DDR SDRAM Voltage > 2.8v
- Chipset Voltage > 1.6v
- AGP Voltage > 1.5v
CPU Over Temp. Protect > 110*C
Advanced chipset features:
Memory Timings > expert
- Row-active delay > 11
- RAS-to-CAS delay > 3
- Row-precharge delay > 3
- CAS latency Time > 2.5
System BIOS cacheable > disabled
Video RAM cacheable > disabled
AGP Aperture Size > 128
FSB Spread Spectrum > disabled
AGP Spread Spectrum > disabled
CPU Thermal-Throttling > disabled
Enhance PCI Performance > Enabled
CPU Disconnect Function > disabled
AGP Data transfer Rate > 8X (Max rate according to card)
AGP Fast Write Capability > disabled
Now save and reboot.
If the PC starts a siren sound when it attempts to POST don't worry it just means you have the 10/10.5x problem. So turn off the PC using the main switch on the PSU or pull out the Power cable if your psu has no switch. Then repower upthe PC and that will have defaulted to safeguard settings. If it still wont POST try again and hold INSERT on boot. If again it won't clear the CMOS as a last resort.
Ok if you have the problem goto
http://www.motherboardfaqs.com and download d_10.bin for the NF7 V2.0 (MAKE SURE ITS FOR THE V2.0 BOARD!).
Using a flashmenu version lower than 1.3 (ie. 1.2) flash to d_10.bin using the "from file" option.
Once succesfully flashed retry the above and behold your new FSB possibilities
If your board hates this bios then you are one of the unfortunate few who haven't got a work around. If you are try d_18.bin as some have got the 10 and 10.5 to work on that.
---------------------------------------------------------------------------------
POINT 2 - The 220 FSB wall with a tbred
If you have a thoroughbred chip and you have done the above and still can't over 220MHz FSB then check out the L12 Mod here:
http://www.geocities.com/trats102002/nf7sFsbsense.html
This should solve your problems as it did me when i had a 2100+ AIUHB chip.
---------------------------------------------------------------------------------
POINT 3 - CPU Interface issues
Now some of you may or may not know but CPU interface is not exacly what it sounds like or describes.
From my testing i have come to the conclusion that CPU-Interface affects both memory performance and stability.
As a quick test and example for you set your system to default CPU speed and FSB and disable CPU-Interface. Boot up, open Sis-Soft Sandra and run memory benchmark, note the results and reboot.
Go back to the bios, enable it and reboot (don't touch anything else) Boot up, open Sis-Soft Sandra and run memory benchmark. Yes, WOW, exactly, your memory scores higher even though its at the same speed
So for those of you crying "Man this memory is crap" it may not be crap just not exceptional as having CPU interface on makes it faster at the same clocks as other boards as it pushes more out of the memory.
---------------------------------------------------------------------------------
POINT 4 - Differenent "better" bios's
From my experience the best bios versions for my board have been:
d_10.bin:
Pros:
- No 10/10.5x multiplier & CPU interface problems
- Highest FSB attained with this bios
Cons:
- High temp (overeads cpu temp)
- Sata corruption problems
d_14beta:
Pros:
- Best clock to clock performance
- No SATA issues
- Highest CPU Speed attained with bios
Cons:
- 10/10.5x multiplier & CPU interface problems
- Beta bios and therefore scary, right?
- Low temp readings, innacurate and worringly low.
Both bios's can be downloaded from
http://www.motherboardfaqs.com (MAKE SURE ITS FOR THE V2.0 BOARD!).
---------------------------------------------------------------------------------
POINT 5 - Vcore / vdimm and +3.3v rail requirements
Vcore:
The NF7 has bad voltage regualation when it comes to vcore. When set at a any level the board under volts by 0.01~0.03v.
For example, to use a true 1.65v the bios needs to be set to 1.675v. The problem becomes even worse with vcores over 1.85v.
This is not as some may assume always a PSU issue as even with a top end PSU with 33AMPS on the +12 rail it still happens.
Also when using high vcore (ie. over 2.0v) with high FSB (ie. over 235) you may encounter posting problems.
Vdimm and the +3.3v rail:
On the NF7 board the vdimm overvolts period. Thats a fact. Despite what the bios shows the H/W monitor on the NF7 is dreadful and way out on vdimm/+3.3v readings. Only way is to use a multimeter. This however is very dangerous as one slip could frag your board.
However i'll make it easy for you, as it is the same with almost all NF7's:
2.9v bios = 3.01v
2.8v bios = 2.90v
2.7v bios = 2.80v
2.6v bios = 2.70v
The +3.3v rail is however hard to determine as it can only be verified by checking the board. As my bios always shows the +3.3 to be 3.33v when itis ofen at the highest 3.51v and at the lowest 3.26v (when using prime95 torture at 3.20v or more vdimm)
The +3.3v rail is only really important at high vdimm usually over 3.0 volts.
For example my vmodded board has it set to roughly 3.25v using a 1K Variable resistor.
If the +3.3v rail is at 3.41 my vdimm is 3.28v, when the rail is at 3.30v it drops to 3.23v and when at as low as 3.26v the vdimm drops down to 3.20v. This is because the vdimm is dependant on how much voltage it can draw from the +3.3v rail.
---------------------------------------------------------------------------------
POINT 6 - How long to run prime and what to look for
Prime95 takes a long time thats the trouble. However for quick FSB/CPU hunting this is what i do.
I'm using version 2.3.7 by the way.
All i do is click torture test select "Blend" and OK it. It then runs iterations of the 1024K primes. If it runs until it says "1024K Passed Sucessful" or something like that, this takes around 15~20mins, i consider that a possiblity for a fully stable FSB / CPU speed.
Once it fails those either up the volts or back off the FSB. Once your happy let it run at that setting for an hour. Usually if it runs for an hour its stable enough for games and general folding etc. You may however encounter a reboot or bsod but unlikely.
If you let it run and i goes over 3hrs i'd consider it rock solid, if it runs upto 8hrs its a boulder, if it runs over a day its mount fuji with a cherry on top, in other words it wont budge.
Also have a read at this how to set up prime correctly:
http://www.ocforums.com/vb/vb/showthread.php?s=&threadid=248225
However, heheh you knew this was coming
, this does not guarantee its 3D stable. It most likely is, but some cards my be unhappy with massive FSB. So i suggest looping 3D mark 2001 over and over to ensure its ok. If it kicks out in Lobby test back down on the FSB until it doesn't.
---------------------------------------------------------------------------------
POINT 7 - How to recover from potential non POST'ing
If at any time your system refuses to boot or makes a siren sound when attempting to post turn off the PC using the main switch on the PSU or pull out the Power cable if your psu has no switch. Then re-power up the PC and that will have defaulted to safeguard settings. If it still wont POST try again and hold INSERT on boot. If again it won't clear the CMOS as a last resort.
If it still wont boot, leave it for a few hours. If still no luck RMA time
---------------------------------------------------------------------------------
POINT 8 - Memory/timing and restrictions.
Now on your hunt for FSB you may encounter some restrictions and no matter what voltage you use it refuses, even a kick in the dimm solves nothing
Well see that advance chipset features section in the bios? get your ass in there. And set as follows.
Memory Timings > expert
- Row-active delay > 11
- RAS-to-CAS delay > 3
- Row-precharge delay > 3
- CAS latency Time > 2.5
Try again and you may get it to stay stable. If you still don't a boot up then drop the CAS from 2.5 to 3.0. If you still don't boot add some more voltage, 2.8 is a general warranted maximum but 2.9v (3.01v actual) prolly wont cause any harm. Ive used 3.38v for days for gods sake lol.
If you have maxed your volts and no luck memory is utter poo, jk.
Also note if your using BH5 memory which hates CAS3.0 and sometimes even CAS2.5.
Either way if you get no luck its time to grab a solder iron and some balls. Yep vmod time.
This mod can be found at
http://www.motherboardfaqs.com
Also while im at it you may have noticed me using a "Row-active delay" or "TRAS" as its usually known as of 11. Yes 11, mmkay?
After testing, its proven that not only is tras 11 more stable, its actually faster in sandra memory bench and 3D Mark 2001
Heres some results from my testing:
*NF7-S @240FSBx10 (480MHz DDR) 6/6 divider and Dual Ch. Sisoft sandra max 3 bandwidth test*
2-2-2-3 Int: 3661(95% eff.)
2-2-2-3 Float: 3431(89% eff.)
2-2-2-4 Int: 3665(95% eff.)
2-2-2-4 Float: 3430(89% eff.)
2-2-2-5 Int: 3666(95% eff.)
2-2-2-5 Float: 3443(89% eff.)
2-2-2-6 Int: 3675(95% eff.)
2-2-2-6 Float: 3440(89% eff.)
2-2-2-7 Int: 3679(95% eff.)
2-2-2-7 Float: 3446(89% eff.)
2-2-2-8 Int: 3674(95% eff.)
2-2-2-8 Float: 3445(89% eff.)
2-2-2-9 Int: 3686(96% eff.)
2-2-2-9 Float: 3448(89% eff.)
2-2-2-10 Int: 3691(96% eff.)
2-2-2-10 Float: 3452(90% eff.)
2-2-2-11 Int: 3702(96% eff.)
2-2-2-11 Float: 3457(90% eff.)
2-2-2-12 Int: 3698(96% eff.)
2-2-2-12 Float: 3450(89% eff.)
Any higher continued to drop.
---------------------------------------------------------------------------------
So there you go, theres my addition to this thread. Hope it helps and live long and prosper etc
Ricjax99 out.
PS. Also note how meaningless a post count/forum status is, not bad for a newb eh? LOL
